To the casual observer of national politics, myself included, Trump’s win in 2016 was surprising. Among serious pundits, Nate Silver was the reliably cooler head, giving Trump better odds than most. But still, Trump’s odds among the odds-makers weren’t good, and the result caught many really good analysts off-guard.
Reasonable people can disagree about why that happened, but our lack of reliable information leading up to a national election plays a role. Most pundits focus on the Electoral College or perhaps polling at the state level. Very little information is available at the sub-state level. So for those of us interested in local or regional trends, we’re left with a lot of noise and very little signal, as Silver might say.
But thanks to organizations like the MIT Election Data and Science Lab, we do have a clearer picture now, and after spending the last few weeks looking at the data, we shouldn’t have been so surprised.
In the last six presidential elections, Republicans have averaged more votes than Democrats in 2,500 counties, or about 80% of all counties (where data is available in the MIT database). Here’s the ranked breakdown:
- Trump 2016: 2,650 counties, where R votes > D votes (84% of total counties)
- Trump 2020: 2,596 (82%)
- Bush 2004: 2,566 (81%)
- Bush 2000: 2,477 (79%)
- Romney 2012: 2,444 (78%)
- McCain 2008: 2,270 (72%)
Same idea here, except this shows where Republicans received more than 50% of the total vote, versus just more votes than Democrats–i.e. this factors in the impact of votes cast for third-party candidates.
- Trump 2020: 2,564 counties, where R votes > 50% of total votes (81% of total counties)
- Bush 2004: 2,531 (80%)
- Trump 2016: 2,523 (80%)
- Romney 2012: 2,373 (75%)
- Bush 2000: 2,308 (73%)
- McCain 2008: 2,197 (70%)
In 2016, there was some separation in terms of Trump performing better head-to-head with the Democratic candidate compared to other recent Republican candidates. However, Trump won roughly the same number of counties outright in 2016 as Bush did in 2004, and improved only slightly on his total in 2020.
So, was there something particularly surprising about the results in 2016–looking strictly at partisan trends and not perceptions of the candidates themselves or any other aspects of the political narrative–or were we just returning to the trendline after Obama’s run? In other words, which election was the anomaly, 2016 or 2008?
After poring over the data the past few weeks, I think we made too much of Trump winning. Clearly, in 2016 and 2020, Trump rode the wave of growing partisanship in counties electing Republican candidates for president. And while he did outperform his Republican predecessors in 2008 and 2012 in terms of the number of counties won, his ability to flip counties from Democrat to Republican was somewhat underwhelming.
Here’s the ranked breakdown of flips from one party to the other in terms of relative votes:
- Obama 2008: flipped 339 counties from R votes > D votes in 2004 to D votes > R votes in 2008
- Trump 2016: 227
- Romney 2012: 200
- Bush 2004: 155
- Biden 2020: 69
- Kerry 2004: 66
- McCain 2008: 43
- Obama 2012: 26
- Clinton 2016: 21
- Trump 2020: 15
On a ranked, net-Republican basis (R flips – D flips), that looks like this:
- Trump 2016: +206 counties
- Romney 2012: +174
- Bush 2004: +89
- Trump 2020: -54
- McCain 2008: -296
Of the counties Trump flipped in 2016, Bush won nearly one-half of them in 2004. Bush also flipped far more counties in his re-election year than Trump did–and had a positive net-flip rate. Further, given the overall trend in Republican-leaning counties since 2000, it’s reasonable to argue that Trump improved only marginally from Romney in 2012, but enough to win.
Pollsters earn a good living working for candidates to track local trends leading up to, and after an election. It’s expensive and provides a competitive advantage. I just wish more of that data was available publicly, somehow. We could use more objective information these days.